通过增强交叉注意力键值，解决复杂多实例图像生成中概念遗漏问题。
Delta-K: Boosting Multi-Instance Generation via Cross-Attention Augmentation
- 在共享交叉注意力键空间中注入语义差分信号，引导生成过程
- 无需训练或修改模型，在多种架构上显著提升组合一致性
- 适合需要高精度多对象生成的视觉设计与内容创作场景
尽管扩散模型在文本到图像生成中表现优异，但在生成复杂多实例场景时常出现概念遗漏。现有无训练方法通过缩放注意力图来解决，却仅加剧了无结构噪声，未能建立连贯语义表征。为此，我们提出Delta-K，一种不依赖主干网络、可即插即用的推理框架，直接在共享交叉注意力键空间中操作。利用视觉语言模型提取编码缺失概念语义特征的差分键ΔK，该信号在扩散过程早期语义规划阶段注入。通过动态优化调度机制，Delta-K将随机噪声锚定为稳定结构，同时保留已有概念。大量实验表明，该方法具有强泛化性：在现代DiT模型和经典U-Net架构上均一致提升组合对齐效果，且无需空间掩码、额外训练或架构修改。
原文摘要 · Abstract (English)
While Diffusion Models excel in text-to-image synthesis, they often suffer from concept omission when synthesizing complex multi-instance scenes. Existing training-free methods attempt to resolve this by rescaling attention maps, which merely exacerbates unstructured noise without establishing coherent semantic representations. To address this, we propose Delta-K, a backbone-agnostic and plug-and-play inference framework that tackles omission by operating directly in the shared cross-attention Key space. Specifically, with Vision-language model, we extract a differential key $ΔK$ that encodes the semantic signature of missing concepts. This signal is then injected during the early semantic planning stage of the diffusion process. Governed by a dynamically optimized scheduling mechanism, Delta-K grounds diffuse noise into stable structural anchors while preserving existing concepts. Extensive experiments demonstrate the generality of our approach: Delta-K consistently improves compositional alignment across both modern DiT models and classical U-Net architectures, without requiring spatial masks, additional training, or architectural modifications.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。